This study investigated the predictive effects of science self-efficacy, value in learning science, science attitudes, gender, and race on science achievement, among United States middle school students utilising Trends in International Math and Science Study 2011 data. A series of preliminary analyses using several descriptive statistics, Cronbach alpha, and exploratory factor analysis, were conducted. The main analysis of the study was Structural Equation Modelling (SEM), utilising both measurement models and structural models. Findings indicated that science self-efficacy, value in learning science, gender and race were significant positive predictors of science achievement, while attitude was not a predictor. Male students showed higher science achievement, science self-efficacy and science attitude than females. White and Asian students showed significantly higher science achievement, science self-efficacy, value in science learning, and science attitudes than Black and Hispanic students. These findings may be utilised to implement new science programs and encourage and train educators.